how do i get clusters from a pointcloud?

i have a point cloud from an object i measured with a 3d microscope.

i am dealing with 3 vectors: x,y containing coordinate points in regular intervals and z containing the z-value (if there is no "Point, resp. no surface" the value is "NaN")

i would like to store and count "clusters" in 2D. By cluster i mean fibres. See pictures from scatter plot below:

how can i handle this? is it easier to convert the data to an image file and use matlabs morphological tools? i would be glad to avoid doing this.. thanks in advance

Answer 1

[Lorenz]

I have exactly the same problem. Did you find a solution? Thanks!

[Stephan]

Yes, I did. The algorithm I used is called DBSCAN (Density-based spatial clustering of applications with noise). Have a look at the code:

http://www.chemometria.us.edu.pl/download/DBSCAN.M

For "eps" I used the diagonal raster-distance of 2 points (sqrt(raster_x^2+raster_y^2)). good luck

[Lorenz]

Great, thanks! I'll have a look at it.